Analysis of Embryonic Malformations in Zebrafish Larvae

Zebrafish larvae has been widely used in testing the effect of chemical substances on the human body. Biological experts have been carrying out these studies manually with a relatively poor productivity. This paper proposes a high-throughput computer vision system to identify and classify embryonic malformations in zebrafish larva. The proposed system, which comprises at least 96-well plates, utilizes multi-classification features including shape, intensity, and texture features. A classifier based on artificial neural networks is utilized to address overlapping problems between different classes of the image dataset and thus improving the classification throughput and accuracy. The preliminary results of this pilot study showed a good performance and low recognition error of the classifier in which two classes were recognized according to the rounding shape of the larva.